Merge pull request #2469 from RosettaCommons/vmullig/multithreaded_peppredict_bg
Get multi-threaded peptide structure prediction working on the "Mira" Blue Gene/Q supercomputer
Tasks:
- [x] Tweak BG/Q build settings.
- [x] Fix a misuse of `delete` in `protocols/noesy_assign/PeakAssignmentParameters.cc`. Note that this was also preventing compilation on Mac/clang.
- [x] Beauty.
- [x] Test compilation on "Mira" Blue Gene/Q supercomputer.
- [x] Benchmark on "Mira". -- Yes, this does allow 2- to 3-fold more sampling than was possible before. This still falls short of the 8-fold improvement that would be the theoretical maximum, though. There's still efficiency work to be done to reduce thread concurrency issues.
Note that this pull request was branched off of vmullig/multithreaded_peppredict, and must therefore be merged _after_ pull request #2430.